{"id":67614,"date":"2026-05-05T20:52:41","date_gmt":"2026-05-05T18:52:41","guid":{"rendered":"https:\/\/www.apfelpatient.de\/?p=67614"},"modified":"2026-05-05T20:53:52","modified_gmt":"2026-05-05T18:53:52","slug":"ios-27-extensions-ai-model-selection-gemini-claude","status":"publish","type":"post","link":"https:\/\/www.apfelpatient.de\/en\/rumors\/ios-27-extensions-ki-modell-wahl-gemini-claude","title":{"rendered":"iOS 27 lets you choose between Gemini, Claude, and more"},"content":{"rendered":"

Apple is opening up its AI system to third-party developers like never before. With iOS 27, users will be able to freely choose which AI model powers Siri, the writing tools, or Image Playground \u2013 including a unique voice for each model. This is according to Bloomberg, citing internal Apple plans.<\/strong><\/p>\n\n\n\n

Apple's AI strategy has been taking shape in several stages for months: The Gemini deal with Google, intended to power the next generation of Siri, was officially confirmed in early 2026<\/a>, while reports of an extension system designed to integrate other AI models have been intensifying. This system has now been given an official name and is significantly larger than previously thought. Apple is calling it \"Extensions\"\u2014and it goes beyond Siri, encompassing virtually the entire Apple intelligence universe. This is according to Bloomberg<\/a>, citing sources familiar with the plans.<\/p>\n\n\n\n

What extensions can specifically do<\/h2>\n\n\n\n

In internal test versions of iOS 27, Apple describes the system as follows: Extensions allow users to access generative AI functions from installed apps on demand \u2013 via Apple Intelligence components such as Siri, the writing tools, Image Playground, and more. In practical terms, this means that users who install the Claude app from Anthropic or the Gemini app from Google's App Store will be able to designate these models as the AI source for native iPhone functions.<\/p>\n\n\n\n

Specifically, this means that when composing text using the writing tools, Apple's own model or Gemini is not mandatory. Users who prefer Claude can select it as the default. The same applies to Image Playground. Siri can also redirect more complex queries to the chosen third-party model \u2013 similar to ChatGPT, but with genuine freedom of choice between multiple providers.<\/p>\n\n\n\n

Different voices for different models<\/h2>\n\n\n\n

Bloomberg highlights a detail that reveals Apple's meticulous attention to detail in a new way: Siri will soon be able to use different voices depending on the AI model in use. Questions answered by the Apple system itself will have one voice, while responses from, for example, Anthropic's Claude will have another. This allows users to hear when their own AI is responding and when an external one is \u2013 a subtle but important element of transparency.<\/p>\n\n\n\n

For Apple, this distinction isn't just an acoustic gimmick. It solves a conceptual problem: Until now, Siri was a single persona. With extensions, Siri becomes the central hub for multiple AI backends, and the voice audibly indicates to the user who they are speaking to. This is, not least, a decision relevant to data privacy and trust.<\/p>\n\n\n\n

What has changed compared to the March report?<\/h2>\n\n\n\n

As early as the end of March 2026, Bloomberg reported on an extension system for Siri<\/a> that would integrate third-party AI apps into the voice assistant. The current report significantly expands on this: Extensions will not only work within Siri, but across the entire Apple intelligence architecture. This includes writing tools, Image Playground, summary functions in Mail and Notes, and other areas of use.<\/p>\n\n\n\n

This gives the system a new dimension. Instead of an AI extension for the voice assistant, Apple is creating a platform-wide interface to which external models can connect. For developers, this means that anyone wanting to score points with a powerful AI model can now integrate it not only into their own app, but also into the system functions that Apple users use every day.<\/p>\n\n\n\n

Strategic move ahead of WWDC<\/h2>\n\n\n\n

The timing of this report is no coincidence. iOS 27 will be unveiled at WWDC 2026 on June 8th<\/a>, and Apple has been visibly focusing its communications on AI topics in the weeks leading up to the event. Extensions are a key component of this strategy: they demonstrate that Apple doesn't intend to compensate for its AI shortcomings with a single partner model, but rather with an open system where users can make choices.<\/p>\n\n\n\n

This is also a communicative response to the criticism of the last two years. Apple has repeatedly been accused of falling behind in the AI race. With Extensions, the company is reformulating its own position: the goal is not to have the best AI model, but rather the best platform on which the best models in the industry come together. Whether this narrative works depends on how smoothly the integration functions in practice.<\/p>\n\n\n\n

Apple Intelligence becomes an open platform<\/h2>\n\n\n\n

With Extensions, Apple is making a shift that is rare in the company's history: Instead of keeping hardware and software as closed as possible, the system is opening up to third-party AI at a deep level. Those who install iOS 27 this fall will likely be able to specify for the first time in the settings which AI model powers the core functions of their iPhone. What this freedom of choice entails will be revealed in the first beta, expected after WWDC. (Image: Shutterstock \/ Tada Images)<\/p>\n\n\n\n